Who holds the federal debt? The pie chart above shows the shares for the last available period:

• 27.1% is held by the U.S. government, its agencies, and its trusts—such as the social security trust.
• 42.1% is held by private individuals and entities in the U.S., which includes 14.2% held by the Federal Reserve. (This 69.2% held domestically is technically debt between Americans.)
• 30.9% is held outside the U.S.

How have these shares evolved over time? The graph below answers this question after removing the inter-agency debt. The Fed’s share of federal debt hasn’t changed much over time. But foreign ownership of debt has: It ramped up in the 1990s and 2000s and has been declining slightly over the past decade.

The last graph shows how these shares translate to a proportion of GDP: The value of debt owed abroad is about a third of annual GDP. The value of debt owed to domestic households and businesses is about a quarter of GDP. For recent years, the lines don’t stack above 100% of GDP, as is often mentioned when talking about the federal debt. The value of debt rises above 100% of GDP only if you include inter-agency debt. And if you also exclude debt held by the Federal Reserve, U.S. federal debt currently amounts to 62% of GDP.

How these graphs were created: For the first graph: Choose the series “Federal Debt Held by Federal Reserve Banks” and “Federal Debt Held by Foreign & International Investors.” Now, to create the series that shows only private domestic holders of federal debt, select “Federal Debt Held by Private Investors” and then use “Add Line” / “Customize data” to include “Federal Debt Held by Foreign & International Investors.” Apply the data transformation a-b. Finally, add a new line after searching for “Federal Debt held by Agencies and Trusts” and divide it by 1000 because it is in different units. Then select graph type “Pie,” which will default to the last observation. For the second graph, go back to the “Edit Graph” format tab and change the graph type to “Area” and stacking to “Percent.” Remove the last series, as it has a shorter sample and makes the percentages jump. Expand the sample period to maximum. For the last graph, use the second graph, but change the stacking to “Normal” and add to each line nominal GDP (make sure not to take real GDP): Divide each line by that series and multiply it by 100 to express it in percentages.

Suggested by Christian Zimmermann.